Hello,
Here is a simple question. I am wondering if there is a way to assign different numbers of processors for different grids in a nested simulation. If not, does it mean that all nested grids have to have the same number of segments?
Thanks!
using MPI for nested simulations
Re: using MPI for nested simulations
currently you need to use the same number of tiles for each grid.
So if you have 12 tiles, then grid 1 could be NtileI=3 and NtileJ = 4; grid 2 could be NtileI=12 NtileJ=1; etc.
but the total has to be 12 (or whatever you use) for each grid.
I am looking into changing this, but it takes a redistribution of the tiles. i am actually thinking to have a way that the child grid is on the same tile as the parent, so that way we dont need to do any mpi gather scatter stuff for the fine2 coarse. it will all be on the same processor.
-j
So if you have 12 tiles, then grid 1 could be NtileI=3 and NtileJ = 4; grid 2 could be NtileI=12 NtileJ=1; etc.
but the total has to be 12 (or whatever you use) for each grid.
I am looking into changing this, but it takes a redistribution of the tiles. i am actually thinking to have a way that the child grid is on the same tile as the parent, so that way we dont need to do any mpi gather scatter stuff for the fine2 coarse. it will all be on the same processor.
-j